NATURAL SELECTION AT MAJOR HISTOCOMPATIBILITY...

P1: KKK/PKS/mbg P2: KKK/plb QC: KKK/arun T1: KKK

September 30, 1998 14:48 Annual Reviews AR069-15

Annu. Rev. Genet. 1998. 32:415–35Copyright c© 1998 by Annual Reviews. All rights reserved

NATURAL SELECTION AT MAJORHISTOCOMPATIBILITY COMPLEXLOCI OF VERTEBRATES

Austin L. Hughes and Meredith YeagerDepartment of Biology and Institute of Molecular Evolutionary Genetics,The Pennsylvania State University, University Park, Pennsylvania 16802;e-mail: [email protected]

KEY WORDS: adaptive evolution, introns, major histocompatibility complex, overdominantselection

ABSTRACT

The loci of the vertebrate major histocompatibility complex encode cell-surfaceglycoproteins that present peptides to T cells. Certain of these loci are highlypolymorphic, and the mechanisms responsible for this polymorphism have beenintensely debated. Four independent lines of evidence support the hypothesis thatMHC polymorphisms are selectively maintained: (a) The distribution of allelicfrequencies does not fit the neutral expectation. (b) The rate of nonsynonymousnucleotide substitution significantly exceeds the rate of synonymous substitutionin the codons encoding the peptide-binding region of the molecule. (c) Polymor-phisms have been maintained for long periods of time (“trans-species polymor-phism”). (d ) Introns have been homogenized relative to exons over evolutionarytime, suggesting that balancing selection acts to maintain diversity in the latter,in contrast to the former.

CONTENTS

INTRODUCTION . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 416

STRUCTURE AND FUNCTION OF MHC MOLECULES. . . . . . . . . . . . . . . . . . . . . . . . . . . 416Class I Molecules. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 416Class II Molecules. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 418

EXPLAINING MHC POLYMORPHISM . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 419The Overdominance Hypothesis. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 419Patterns of Nucleotide Substitution. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 421Alternative Hypotheses. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 423

4150066-4197/98/1215-0415$08.00

Ann

u. R

ev. G

enet

. 199

8.32

:415

-435

. Dow

nloa

ded

from

arj

ourn

als.

annu

alre

view

s.or

gby

b-o

n: U

nive

rsid

ade

do P

orto

(U

P) o

n 02

/10/

09. F

or p

erso

nal u

se o

nly.



416 HUGHES & YEAGER

TRANS-SPECIES POLYMORPHISM. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 425

CLASS I INTRONS. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 428Nucleotide Diversity in Exons and Introns. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 428Hitch-Hiking under Balancing Selection. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 430

INTRODUCTION

The major histocompatibility complex (MHC) of vertebrates is a multigenefamily whose products are cell-surface glycoproteins that play a key role inthe immune system by presenting peptides to T cells (32). The MHC familyincludes two major subfamilies, called class I and class II. In most of the ver-tebrate species in which these genes have been mapped, the class I and class IIfamilies are linked together in a single gene complex. This complex is locatedon chromosome 6 in humans and is called the HLA complex (for “human leuko-cyte antigen”). In mammals, class I and class II genes are located in differentregions of this complex, which are separated by a third region, sometimes calledclass III, that contains unrelated genes. Certain of the class I and class II geneshave extraordinarily high levels of polymorphism, among the highest knownin any organism (32). Furthermore, MHC polymorphisms are characterized bya large number of alleles of intermediate frequencies—a pattern of polymor-phism inconsistent with selective neutrality but rather suggesting the action ofsome form of balancing selection (21).

To allow readers unfamilar with the MHC or with immunology to appre-ciate our discussion of the molecular evolution of MHC genes, we begin thisreview with an introduction to basic MHC biology. Then we discuss evidencethat MHC polymorphism is maintained by balancing selection relating to thepeptide-binding function of the MHC molecules and thus, ultimately, to diseaseresistance. Finally, we consider some unique aspects of MHC evolution thatare ultimately consequences of this balancing selection.

STRUCTURE AND FUNCTION OF MHC MOLECULES

Class I MoleculesThe polymorphic class I MHC molecules (called the class Ia molecules or class Iclassical molecules) are glycoproteins expressed on the surface of all nucleatedsomatic cells; they function to present peptides to cytotoxic T lymphocytes(CTL). The class I molecule is a heterodimer consisting of the following twochains: (a) anα chain or heavy chain, made up of three extracellular domains(designatedα1,α2, andα3), a transmembrane region, and a cytoplasmic domain;and (b) a molecule calledβ2-microglobulin (β2m), which consists of a singledomain (Figure 1).β2m is noncovalently linked to theα3 domain. Theα chains

Ann

u. R

ev. G

enet

. 199

8.32

:415

-435

. Dow

nloa

ded

from

arj

ourn

als.

annu

alre

view

s.or

gby

b-o

n: U

nive

rsid

ade

do P

orto

(U

P) o

n 02

/10/

09. F

or p

erso

nal u

se o

nly.



MAJOR HISTOCOMPATIBILITY COMPLEX LOCI 417

Figure 1 Schematic illustrations of MHC class I and class II molecular structure. PBR, peptidebinding region.

are encoded within the MHC complex by the polymorphic class Ia loci, ofwhich there are three in humans (HLA-A,HLA-B, andHLA-C). In mammals andprobably in most other vertebrates,β2m is encoded outside the MHC complex(on chromosome 15 in humans).β2m shows evidence of a distant evolutionaryrelationship to class Iα chains and to class II MHC molecules, but the locusencoding it is not polymorphic.

In all cells, there is a constant turnover of cellular proteins that are brokendown into small peptides by a multimeric proteolytic complex in the cytoplasmknown as the proteasome (46, 50). Proteasomes are present in all organisms,but their components are considerably more highly diversified in eukaryotesthan they are in archaebacteria (22, 50). In mammals (and probably in mostother vertebrates), there are two proteasome components encoded within theMHC class II region, called LMP2 and LMP7. (LMP is an abbreviation forlow molecular mass polypeptide.) LMP2 and LMP7 are not expressed underall circumstances. Rather, two constitutively expressed components, called Xand Y, are ordinarily expressed in their place (2). The cytokineγ -interferonenhances expression of both class I MHC molecules and LMP2 and LMP7. Aproteasome containing LMP2 and LMP7 (called an LMP+ proteasome) hasan altered specificity with regard to where it cleaves polypeptides; the LMP+proteasome specifically produces peptides of a sort likely to be bound by class IMHC molecules (14, 17, 39).

These peptides are transported across the membrane of the endoplasmicreticulum (ER) by a dimeric transporter called TAP. The two subunits of TAPare themselves encoded in the MHC class II region. In the ER, a complex is

Ann

u. R

ev. G

enet

. 199

8.32

:415

-435

. Dow

nloa

ded

from

arj

ourn

als.

annu

alre

view

s.or

gby

b-o

n: U

nive

rsid

ade

do P

orto

(U

P) o

n 02

/10/

09. F

or p

erso

nal u

se o

nly.



418 HUGHES & YEAGER

formed involving the class I MHC molecule, the peptide, andβ2m, which isthen transported to the cell surface. When the first crystal structure of a classI MHC molecule was described, its most striking feature was a groove at thetop of the molecule formed by twoα helices bordering aβ-pleated sheet (6, 7).Residues from both theα1 andα2 domains contribute to this groove (Figure 1).It seemed obvious that this groove was where the peptide is bound, a hypoth-esis later confirmed by crystallographic images of class I molecule complexedwith peptides (18, 53). The class I peptide-binding region (PBR) consists offive pockets (pockets A–F) in which side-chains of the peptide residues fit(52).

In an uninfected cell, the peptides bound by a class I molecule are derivedfrom the cell’s own proteins (often called self peptides.). CTL exercise a con-tinual surveillance in the body by means of their cell-surface receptors (T cellreceptors or TCR). In the development of CTL in the thymus, TCR are selectedso that the only CTL permitted to circulate are those that do not attack thecomplex of self class I MHC and self peptide (5). However, during infection bya virus or other intracellular parasite, some of the proteins broken down by theproteasome are of parasitic origin. Thus at least some of the class I moleculesexpressed on the surface of an infected cell will bind nonself or foreign peptides.When CTL encounter the complex of self class I MHC and foreign peptide, acytotoxic reaction is initiated that kills the infected cell. CTL can only recog-nize foreign peptides in the context of self class I MHC; this phenomenon isknown as class I MHC restriction of CTL. The CTL and class I MHC togetherthus provide a drastic solution to the problem of an intracellular parasite: killingall cells that harbor the infection (5). Mice that do not expressβ2m and, thus,do not express class I MHC on their cell surfaces, suffer severe effects whenexposed to intracellular pathogens. For example, these mice showed delayedviral clearance and increase mortality when infected with influenza virus (3) and100% mortality when infected with the intracellular bacteriumMycobacteriumtuberculosis(16). These results show that the class I MHC plays an essentialrole in immune defense against intracellular pathogens.

Class II MoleculesClass II MHC molecules have a much more restricted expression pattern thando class I molecules, in that they are expressed primarily on antigen-presentingcells of the immune system. The class II molecule presents peptides to helperT cells. In response to a foreign peptide, the helper T cells release cytokinesthat trigger an appropriate immune response (including the production of an-tibodies). The class II molecule is similar to the class I molecule in havingfour extracellular domains, but it achieves this structure in a rather differentway (Figure 1). The class II molecule is a heterodimer consisting of anα chain

Ann

u. R

ev. G

enet

. 199

8.32

:415

-435

. Dow

nloa

ded

from

arj

ourn

als.

annu

alre

view

s.or

gby

b-o

n: U

nive

rsid

ade

do P

orto

(U

P) o

n 02

/10/

09. F

or p

erso

nal u

se o

nly.




and aβ chain, each of which in mammals is encoded in the class II region ofthe MHC complex. In placental mammals, the class II region is divided intosubregions (designated DR, DP, and DQ in humans), each of which contains afunctionalα chain gene and one or more functionalβ chain genes. Theα chainincludes two extracellular domains (α1 andα2), a transmembrane region, anda cytoplasmic tail.

Like the class I molecule, the class II molecule binds peptides in a grooveat the top of the molecule. As with class I, the class II peptide-binding grooveconsists of twoα helices bordering aβ-pleated sheet. The difference is that inclass II, one of theα helices and about half of theβ-pleated sheet are contributedby theα chain, whereas the otherα helix and the other half of theβ-pleatedsheet come from theβ chain (Figure 1). Unlike the peptides presented by class I,which are mainly 9 amino acids in length, the peptides presented by class IImolecules can vary substantially in length, between about 11 and 17 residues(49). The reason for this difference is that in the case of class I the ends of thepeptide are tucked down into the peptide-binding groove, limiting the peptide’slength. In class II, the peptide’s ends are free, which makes its length lessconstrained.

The complex between the class II molecule and its peptide ligand is createdby a mechanism quite distinct from that of class I. Before transport to thecell surface, the class II dimer forms a complex with a polypeptide known asthe invariant chain (Ii). This complex then travels to an acidic endosome-likecompartment (47). There Ii is degraded, and the class II molecule binds thepeptide which it transports to the cell surface. A molecule known as DM servesas a chaperone facilitating the loading of peptides by class II molecules (34).Interestingly, DM is clearly evolutionarily related to the class II molecule; itconsists of anα chain and aβ chain, each of which shows clear evidence ofhomology to the corresponding chains of the class II heterodimer (29).

EXPLAINING MHC POLYMORPHISM

The Overdominance HypothesisZinkernagel & Doherty (59) first demonstrated class I MHC restriction of anti-gen (i.e. peptide) recognition by CTL. Soon afterwards, Doherty & Zinkernagel(13) proposed the first hypothesis for MHC polymorphism that took into ac-count the actual biological function of these molecules. Doherty & Zinkernagelpresented evidence that different class I MHC gene products differ with respectto the antigens that they can present. In other words, to express the concept interms of our current knowledge of MHC function, different allelic products binddifferent arrays of peptides Thus, they argued, in a population exposed to anarray of pathogens, it will be advantageous for an individual to be heterozygous

Ann

u. R

ev. G

enet

. 199

8.32

:415

-435

. Dow

nloa

ded

from

arj

ourn

als.

annu

alre

view

s.or

gby

b-o

n: U

nive

rsid

ade

do P

orto

(U

P) o

n 02

/10/

09. F

or p

erso

nal u

se o

nly.



420 HUGHES & YEAGER

Table 1 Examples of anchor residues (boldface) and auxiliaryanchor residues of peptides bound by the human class I loci MHCHLA-A and HLA-B

Residue position

Allele 1 2 3 4 5 6 7 8 9 (10)a

A1 T D L YS E

A∗0201 L VM L

A68.1 V RT K

B∗ H I39011

R V LL

B∗4403 E YF

B53 P

aThe last residue position of the peptide is usually number 9 but maybe number 10. Data are from Reference 49.

at MHC loci because a heterozygote will be able to present a broader array ofantigens and thus resist a broader array of pathogens. Such a mechanism ofheterozygote advantage (also known as overdominant selection) could accountfor the extraordinary polymorphism found at MHC loci.

Doherty & Zinkernagel’s early evidence pointing to a difference betweendifferent MHC gene products with respect to the peptides they bind has beenconfirmed in recent years by sequencing peptides bound by MHC molecules.Comparisons of many such peptides have shown that the peptides bound bya specific MHC allelic product invariably contain one or more characteristicresidues—called anchor residues, because they anchor the peptide into thebinding groove (Table 1). In the case of the class I MHC, the anchor residues areusually the second residue of the peptide, which fits into the B pocket, and /orthe ninth residue of the peptide, which fits into the F pocket (49). Becausepositions other than the anchor residues seem to be relatively free to vary, eachallelic product can potentially bind thousands of different peptides. Nonetheless,because different allelic products have different anchor motifs, a heterozygotewill presumably have much broader immune surveillance than a homozygotehas. For example, HLA-B∗39011 prefers the positively charged residues Ror H in the second residue of the peptide, whereas HLA-B∗4403 prefers thenegatively charged residue E (Table 1). An individual heterozygous for thesetwo alleles will be able to bind both types of peptides.

Ann

u. R

ev. G

enet

. 199

8.32

:415

-435

. Dow

nloa

ded

from

arj

ourn

als.

annu

alre

view

s.or

gby

b-o

n: U

nive

rsid

ade

do P

orto

(U

P) o

n 02

/10/

09. F

or p

erso

nal u

se o

nly.




Doherty & Zinkernagel’s hypothesis that overdominant selection maintainsMHC polymorphism did not initially meet with wide acceptance. Some pop-ulation geneticists had the mistaken impression that overdominant selectioncannot maintain a polymorphism as extensive as that seen at MHC loci. Thisimpression was based on theoretical models that did not take into account therole of mutation in incorporating new alleles. More realistic models that in-corporated the role of mutation showed that overdominant selection is indeedcapable of maintaining a high level of polymorphism (37).

An additional problem was the difficulty in testing the hypothesis of over-dominant selection at MHC loci by means of a conventional population study.In an outbred species such as human or mouse, most individuals are heterozy-gous at most MHC loci. Thus, it would be necessary to survey many thousandsof individuals to amass a large enough sample of homozygotes to comparetheir fitness with that of heterozygotes. Furthermore, if the selective advantagepossessed by heterozygotes were small—say one or a few percent—then thesample size would have to be still larger to have the statistical power to test fora difference between homozygotes and heterozygotes.

Patterns of Nucleotide SubstitutionHughes & Nei (24) took a different approach to testing Doherty & Zinkernagel’shypothesis. By the late 1980s, a number of nucleotide sequences for MHC classI genes had become available. In most genes the rate of synonymous nucleotidesubstitution per site (dS) exceeds that of nonsynonymous substitution per site(dN). Theoretical study (36) had predicted that overdominant selection shouldenhance the rate of codon substitution; thus, if Doherty & Zinkernagel’s hy-pothesis is true,dN should be enhanced in the case of MHC genes. Furthermore,the first crystal structure of a class I MHC molecule had recently been published(6, 7), revealing the peptide-binding groove. On the hypothesis that MHC poly-morphism is maintained by overdominant selection relating to peptide binding,Hughes & Nei (24) predicted that an enhanced nonsynonymous rate should beseen mainly in the codons encoding the PBR of the molecule.

The results dramatically confirmed this prediction. Figure 2 shows the resultsof recent analyses using many more sequences than were available to Hughes& Nei (24), but the results are essentially the same as they reported. In the57 codons encoding the PBR,dN significantly exceedsdS (Figure 2). By con-trast, in the non-PBR portions of theα1 andα2 domains and in theα3 domain,dS exceedsdN, as is true of most genes (Figure 2). Note thatdS values do notdiffer greatly from one gene region to another. SincedS is expected to reflectthe mutation rate (the fraction of neutral mutations at synonymous sites be-ing close to 100%), the uniform value ofdS indicates that the enhanced valueof dN in the PBR codons cannot be explained by a higher mutation rate in

Ann

u. R

ev. G

enet

. 199

8.32

:415

-435

. Dow

nloa

ded

from

arj

ourn

als.

annu

alre

view

s.or

gby

b-o

n: U

nive

rsid

ade

do P

orto

(U

P) o

n 02

/10/

09. F

or p

erso

nal u

se o

nly.



422 HUGHES & YEAGER

Figure 2 Mean numbers of synonymous (dS) and nonsynonymous (dN) nucleotide substitutionsper site (41), with their standard errors (42), for pairwise comparisons among alleles at theHLA-A(A; 48 alleles).HLA-B (B; 99 alleles), andHLA-C (C; 35 alleles) loci.dS anddN were estimatedseparately for the peptide-binding region (PBR) codons and for the remainder of theα1 andα2domains. Tests of they hypothesis thatdS = dN: ∗P< 0.05;∗∗∗P< 0.001.

those codons. Rather, the results strongly support the hypothesis that positiveDarwinian selection has acted to enhance the rate of nonsynonymous substitu-tion in the PBR codons and thus to enhance amino acid diversity in the PBR.

In the case of the class II MHC, a hypothetical structure was proposed byanalogy with the known class I structure (8). Using this hypothetical structure,Hughes & Nei (25) founddN > dS in the putative PBR. When a class II crystalstructure was obtained (9), this finding was confirmed (Figure 3; 29).

Ann

u. R

ev. G

enet

. 199

8.32

:415

-435

. Dow

nloa

ded

from

arj

ourn

als.

annu

alre

view

s.or

gby

b-o

n: U

nive

rsid

ade

do P

orto

(U

P) o

n 02

/10/

09. F

or p

erso

nal u

se o

nly.




Figure 3 Mean numbers of synonymous (dS) and nonsynonymous (dN) nucleotide substitutionsper site (41), with their standard errors (42), for pairwise comparisons among alleles at theHLA-DRB1 (124 alleles),HLA-DQB1 (14 alleles), andHLA-DPB1 (49 alleles) loci. dS anddN wereestimated separately for the peptide-binding region (PBR) codons and for the remainder of theβ1domain. Tests of the hypothesis thatdS = dN: ∗∗∗P< 0.001.

Alternative HypothesesComparison of rates of synonymous and nonsynonymous nucleotide substitu-tion provided strong evidence that MHC polymorphism is maintained by someform of balancing selection; but it is still uncertain whether this selection isoverdominant, as hypothesized by Doherty & Zinkernagel, or rather representssome other form of balancing selection. One form of balancing selection thathas received a great deal of attention in the literature of theoretical populationgenetics is frequency-dependent selection. Of the several different models of

Ann

u. R

ev. G

enet

. 199

8.32

:415

-435

. Dow

nloa

ded

from

arj

ourn

als.

annu

alre

view

s.or

gby

b-o

n: U

nive

rsid

ade

do P

orto

(U

P) o

n 02

/10/

09. F

or p

erso

nal u

se o

nly.



424 HUGHES & YEAGER

frequency-dependent selection, some are theoretically capable of maintaininga high level of polymorphism such as seen at MHC loci (55). In the case of theMHC, however, there is a clear rationale for overdominant selection based onthe function of the molecules: namely, that heterozygotes have an advantagederived from a broader immune surveillance because a heterozygote can binda broader spectrum of peptides than can a homozygote.

One early hypothesis to explain MHC polymorphism was independent of themolecules’ function. This was the hypothesis that MHC loci have an unusuallyhigh mutation rate (1). DNA sequence data have made it possible to test thishypothesis rigorously. BecausedS is expected to reflect the mutation rate,dSvalues for MHC genes can be compared with those of other genes to assessthe comparative magnitude of the mutation rate at MHC loci. Such compar-isons have shown that the mutation rates at MHC loci are below average formammalian genes. For example, whendS values between human and mousewere computed for 80 immunoglobulin superfamily C-type domains, the meanvalue was 0.654± 0.026 (23). The mean for class IIβ2 domains, which arehomologous to immunoglobulin C-type domains, was 0.526± 0.067 (23).

A more recent version of essentially the same hypothesis held that MHCpolymorphism was enhanced by interlocus recombination (gene conversion)(36, 45). Theoretically, it is possible that if members of a gene family havediverged from each other at the sequence level and interlocus recombinationsubsequently occurs, polymorphism at each locus will be enhanced. However,gene conversion is expected to be an essentially random process; thus it cannotexplain the very specific pattern ofdN> dSin the PBR codons that characterizesMHC loci (24, 25).

Because the function of MHC molecules was unknown for a long time, theearliest hypotheses to explain MHC polymorphism often attempted to explainboth MHC function and polymorphism. Several popular hypotheses relied onanalogies between the MHC and other biological systems. Most influential werehypotheses that saw an analogy between the MHC and the self-incompatibilitysystems of plants. The self-incompatibility loci are also extraordinarily poly-morphic, and it was tempting to see the MHC as a vertebrate analogue. Theresult of this analogy was a proliferation of hypotheses relating the MHC to re-production (e.g. 56). Even though the real function of the MHC is now known,some of these hypotheses have assumed a life of their own in the literature andcontinue to attract adherents.

One hypothesis was that MHC polymorphism is maintained by maternal-fetal interactions (12). This hypothesis depends on the assumption that the pro-duction of maternal antibodies to fetal class I MHC molecules has a beneficialeffect on fetal growth and survival. Although some early studies seemed to showsuch an effect, it has not been supported by subsequent work (11, 30, 32, 57).

Ann

u. R

ev. G

enet

. 199

8.32

:415

-435

. Dow

nloa

ded

from

arj

ourn

als.

annu

alre

view

s.or

gby

b-o

n: U

nive

rsid

ade

do P

orto

(U

P) o

n 02

/10/

09. F

or p

erso

nal u

se o

nly.




Furthermore, it is hard to imagine how maternal-fetal interactions would lead tonatural selection favoring diversity specifically in the PBR. On this hypothesis,one would predict that selection would enhance the rate of nonsynonymoussubstitution in epitopes for maternal antibodies. In the case of class I MHCmolecules, these epitopes are scattered throughout theα1 and α2 domains,rather that being concentrated in the PBR (7, 40). Also, this hypothesis can-not account for the fact that MHC polymorphism is high in fish, amphibians,and birds, all of which lack maternal-fetal interactions. Finally, it cannot ac-count for class II polymorphism since class II molecules are only expressedon antigen-presenting cells of the immune system and thus are unlikely to beinvolved in maternal-fetal interactions.

Another hypothesis is that MHC polymorphism is maintained by disassorta-tive mating on the basis of MHC genotype, which is supposedly recognized viaolfaction. Experiments cited as evidence for this phenomenon in mice (48, 58)lack relevant controls and are open to other interpretations (reviewed in 23).Data from the S-leut Hutterite religious isolate have been presented as evi-dence for MHC-based disassortative mate choice in humans (44). This is anendogamous population descended from a small number of founders, in whichmembers avoid marriage to first cousins. In such a population, avoidance ofclosely consanguineous marriage alone will lead to a lower frequency of sharingof MHC haplotypes by spouses than would be expected under random mating.To test the hypothesis of MHC-based mate choice, it is necessary to comparethe frequency of MHC-haplotype sharing between actual spouses with that be-tween potential spouses (i.e. individuals of the appropriate sex, age, and degreeof kinship to be chosen as spouses). So far, such a comparison has not beenmade (44). Furthermore, no evidence of MHC-associated mate choice was ob-tained in South Amerindians (20), whose population structure is probably moretypical of human populations throughout history than are the highly inbred Hut-terites. As with maternal-fetal incompatibility, it is hard to imagine how MHC-based mate choice would lead to natural selection focused specifically on thePBR.

TRANS-SPECIES POLYMORPHISM

A characteristic of MHC polymorphism that provides additional strong supportfor the hypothesis of balancing selection is the phenomenon called trans-speciespolymorphism. Often MHC polymorphisms are quite ancient, predating spe-ciation events. For example, certain alleles at both class I and class II MHCalleles from human and chimpanzee belong to allelic lineages that have per-sisted since before these two species diverged 5–7 MYA (19, 35, 38). Figure 4shows an example of trans-species polymorphism in the class IIDQB1locus in

Ann

u. R

ev. G

enet

. 199

8.32

:415

-435

. Dow

nloa

ded

from

arj

ourn

als.

annu

alre

view

s.or

gby

b-o

n: U

nive

rsid

ade

do P

orto

(U

P) o

n 02

/10/

09. F

or p

erso

nal u

se o

nly.



426 HUGHES & YEAGER

Ann

u. R

ev. G

enet

. 199

8.32

:415

-435

. Dow

nloa

ded

from

arj

ourn

als.

annu

alre

view

s.or

gby

b-o

n: U

nive

rsid

ade

do P

orto

(U

P) o

n 02

/10/

09. F

or p

erso

nal u

se o

nly.




human and chimpanzee. In the phylogenetic tree ofDQB1alleles, the humanallelesHLA-DQB1∗0302andHLA-DQB1∗03032cluster with the chimpanzeeallelePatr-DQB1∗0302(Figure 4). On the other handHLA-DQB1∗0501, HLA-DQB1∗06011and related human alleles cluster with the chimpanzee allelePatr-DQB1∗O601and related chimpanzee alleles (Figure 4). These clusters ofalleles apparently represent allelic lineages that were present in the commonancestor of chimpanzees and humans and have persisted in each populationsince their divergence.

Neutral polymorphisms are not expected to persist very long in populations.Coalescence theory predicts that, for pairs of neutral alleles selected at randomfrom a locus in a randomly mating population, their mean coalescence time (thatis, the time of their last common ancestor) will be 2Ne generations, whereNe

is the long-term effective population size (55). Assuming a long-term effectivepopulation size of 104 for humans, the mean coalescence time for neutral alleleswould be only 600,000 years. Thus, neutral polymorphism is, with respect toevolutionary time, a relatively transient phenomenon.

Under balancing selection, Takahata & Nei (55) showed that polymorphismscan persist much longer than in the neutral case. Using computer simulation,these authors studied overdominant selection and several models of frequency-dependent selection. They found that under overdominant selection and one typeof frequency-dependent selection (which they called minority advantage), it waspossible to maintain polymorphisms for very long times even with relativelymodest selection coefficients. Thus either of these types of balancing selectioncan account for the long coalescence times of alleles at MHC loci.

In the model of minority advantage, it was assumed that a genotype had aselective advantage whenever it became rare in the population (54). Mathemat-ically, this model turns out to be essentially the same as that of overdominantselection; however, from a biological point of view, it might well be questionedwhether it is truly applicable to the MHC. Consider a parasite species that iseasily eliminated by its host because most members of the host species bear anMHC allele (A1) whose product can bind and present a peptide from a givenprotein of the parasite. Then, suppose that a mutation occurs in the parasite sothat this MHC allele can no longer bind the peptide, and the parasite is now ableto infect most members of the host species with impunity. Suppose, however,

←−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−Figure 4 Phylogenetic tree ofβ1 domain of human (HLA-) and chimpanzee (Patr-) DQB1 se-quences, constructed by the neighbor-joining method (51) based on the proportion of amino aciddifference (p). The tree is rooted with sequences from the orthologousA locus fromMusspecies:M. domesticus(Mudo-),M. spicilegus(Musi-), andM. musculus(Mumu-). The numbers on inter-nal branches represent the percentage of 1000 bootstrap samples (15) supporting that branch; onlyvalues>50% are shown.

Ann

u. R

ev. G

enet

. 199

8.32

:415

-435

. Dow

nloa

ded

from

arj

ourn

als.

annu

alre

view

s.or

gby

b-o

n: U

nive

rsid

ade

do P

orto

(U

P) o

n 02

/10/

09. F

or p

erso

nal u

se o

nly.



428 HUGHES & YEAGER

that there is a rare MHC allele in the host species (A2) whose product can bindanother peptide from this parasite and protect against infection. Clearly, the A2allele will have a selective advantage and will increase in frequency.

Now consider what will happen to the A1 allele. The minority advantagemodel assumes that, once the A2 allele becomes common, the A1 allele willagain have a selective advantage. But how would this happen realistically in thecase of the MHC? It might be that the parasite will mutate again so that the A2allelic product can no longer efficiently bind a peptide from the parasite. But itseems rather unlikely that this new escape mutant will somehow restore bindingby the A1 allelic product. Yet this is precisely what the minority advantagemodel requires. Therefore, this model may not be applicable to the case of theMHC.

On the other hand, after the parasite mutates so that the A2 allelic product nolonger binds a peptide from its proteins, it might happen that still another newmutant MHC allele appears (A3), the product of which can efficiently bind apeptide from the parasite. If this happens, we can expect that the A3 allele willincrease in frequency. This model is called the pathogen adaptation model byTakahata & Nei (55). However, this model cannot explain what is happeningat MHC loci because rather than leading to a long-lasting polymorphism, thisprocess will lead to a turnover of alleles over time (24, 55).

CLASS I INTRONS

Nucleotide Diversity in Exons and IntronsRecently, intron sequences have become available for a number of alleles at thehuman class I loci HLA-A, -B, and -C. The evolutionary dynamics of class Iintrons seems to differ strikingly from that of exons in ways that may seemsurprising to some immunologists. Yet the properties of class I introns are infact exactly what one would predict in the case of balancing selection. Thus,the analysis of class I introns has provided an additional, independent line ofevidence that polymorphism at these loci is maintained by balancing selection.

Figure 5 shows plots of mean proportion of nucleotide different (p) for allpairwise comparisons among alleles at the HLA-A, -B, and -C loci. The regionsof the gene compared are exons 2–3, which encode theα1 andα2 domains,including the PBR codons; exon 4, which encodes the conservedα3 domain;and the first three introns of the gene (introns 1–3). One striking aspect of these

−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−→Figure 5 Proportion of nucleotide difference in a sliding window of 30 base pairs in all pairwisecomparisons among alleles at theHLA-A, -B, and-C loci, from the beginning of intron 1 to the endof exon 4.Horizontal barsindicate the position of exons 2, 3, and 4. Reprinted from Reference 10.

Ann

u. R

ev. G

enet

. 199

8.32

:415

-435

. Dow

nloa

ded

from

arj

ourn

als.

annu

alre

view

s.or

gby

b-o

n: U

nive

rsid

ade

do P

orto

(U

P) o

n 02

/10/

09. F

or p

erso

nal u

se o

nly.




Ann

u. R

ev. G

enet

. 199

8.32

:415

-435

. Dow

nloa

ded

from

arj

ourn

als.

annu

alre

view

s.or

gby

b-o

n: U

nive

rsid

ade

do P

orto

(U

P) o

n 02

/10/

09. F

or p

erso

nal u

se o

nly.



430 HUGHES & YEAGER

Table 2 Number of nucleotide substitutions per 100 sites (d ) in introns 1, 2, and 3 and per 100synonymous sites (dS) in exons 2–3 for comparisons among HLA class I alleles

d dSComparison Intron 1 Intron 2 Intron 3 Exons 2–3

Means for all HLA-A locus 4.2± 1.2 2.3± 0.6 2.2± 0.4 3.5± 0.8pairwise HLA-B locus 2.5± 0.9 1.6± 0.4∗∗∗ 0.7 ± 0.2∗∗∗ 4.9 ± 0.9comparisons HLA-C locus 1.6± 0.6 1.8± 0.5 1.4± 0.3∗∗ 3.6 ± 0.9(intralocus)a

All intralocus 2.7± 0.9 1.8± 0.5∗∗∗ 1.1 ± 0.3∗∗∗ 4.4 ± 0.9

Selected A∗1101vs 0.8± 0.8∗∗ 2.8 ± 1.0∗ 0.7 ± 0.4∗∗ 9.3 ± 2.8individual A∗3002comparisons

A∗2501vs 0.0± 0.0 0.0± 0.0 2.6± 0.7∗∗∗ 0.0 ± 0.0A∗2601

B∗0702 vs 0.0± 0.0 1.3± 0.7 0.7± 0.4 2.7± 1.4B∗4201

B∗0702 vs 5.0± 0.2 0.0± 0.0 1.5± 0.5 6.8± 2.3B∗5401

Cw∗0602 vs 1.6± 1.2 1.7± 0.9 0.4± 0.3∗ 4.6 ± 1.9Cw∗1203

aNumbers of alleles compared for each locus are as follows: HLA-A, 15; HLA-B, 23; HLA-C, 12. Tests ofthe hypothesis thatd in an intron equalsdS in exons 2–3:∗P< 0.05;∗∗P< 0.01;∗∗∗P< 0.001.

plots is that the meanp is generally lower in the introns than in the exons; thisis particularly true of intron 3, which is much longer than either intron 1 orintron 2 (Figure 5). In most genes, this pattern would be reversed:p wouldbe lower in exons than in introns because purifying selection eliminates mostnonsynonymous mutations in exons (28).

A detailed examination of patterns of nucleotide substitution explains theusual results of the sliding window analysis. At each locus, the mean numberof nucleotide substitutions per site (d ) in introns 1–3 was compared with meandS in exons 2–3 (Table 2). MeandS in the exons was generally higher than meand in the introns. This was particularly true for intron 3 and was most strikingin the case of theB locus. At theB locus, meand in intron 3 was less than1%, whereas meandS in exons 2–3 was nearly 5% and seven times higher thanmeand in intron 3 (Table 2). This result is very unusual because in the case ofmost genesd in introns anddS in exons are about equal (28).

Hitch-Hiking under Balancing SelectionThe most reasonable explanation for the fact thatd in introns of human classI genes is often lower thandS in exons is that introns are homogenized by

Ann

u. R

ev. G

enet

. 199

8.32

:415

-435

. Dow

nloa

ded

from

arj

ourn

als.

annu

alre

view

s.or

gby

b-o

n: U

nive

rsid

ade

do P

orto

(U

P) o

n 02

/10/

09. F

or p

erso

nal u

se o

nly.




interallelic recombination and subsequent genetic drift (10). The exons ofthese genes—particularly those encoding the PBR—are quite ancient, havingbeen maintained for millions of years by balancing selection. However, thisselection does not apply to introns. Though intron sequences may hitch-hikealong with exon sequences to some extent, if recombination and drift lead toloss of ancient polymorphism in an intron, this will be selectively neutral. Thus,introns of MHC genes are expected to be evolutionarily younger on averagethan are the exons encoding the PBR. BothdS in exons andd in introns areexpected to reflect the mutation rate, since most mutations at synonymous sitesand at sites in introns are selectively neutral (28). WhendS in the exons is muchhigher thand in adjacent introns, the most straightforward interpretation is thatthe exons are older than the introns.

Population geneticists have extensively studied the problem of polymorphismat a locus linked to one under balancing selection (33, 43, 54). These studiespredict that the degree of hitch-hiking—and thus the extent of polymorphism—at such a locus will be a function of the extent of recombination between thatlocus and the one under selection. Extending these predictions to the case ofpolymorphic class I MHC loci, we expect that introns more closely linked toexons 2–3 (encoding the PBR) will show higher levels of polymorphism thanthose less closely linked to exons 2–3. Introns 2 and 3 are relatively short (130and 268 aligned nucleotide sites, respectively); and intron 1 is located just 5′ toexon 2, while intron 2 is located between exons 2 and 3. By contrast, intron 3,located 3′ to exon 3, contains 653 aligned sites; so, on average, nucleotide sitesin intron 3 will be less closely linked to nonsynonymous PBR sites than willsites in introns 1 and 2. Thus, we might predict that intron 3 will be more likelyto be homogenized by recombination and subsequent drift than will introns1–2; the latter two introns are predicted to hitch-hike more closely with thePBR exons and thus to have higher levels of nucleotide diversity.

These predictions are supported by the data (Table 2). Intron 3 sequencesshow the lowest meand for all three loci (Table 2). In addition, comparisonsbetween individual sequences reveal some apparent recent cases of recombi-nation. For example, the allelesA∗2501andA∗2601are identical in exons 2–3and in introns 1 and 2, yet differ markedly in intron 3 (Table 2), which suggeststhat that this intron has been recently donated to one of these alleles by a moredistantly related allele. On the other hand,B∗0702andB∗5401, though identicalin intron 2, are highly divergent in the remainder of their sequence (Table 2). Inthis case, recombination seems to have caused the homogenization of intron 2between these two alleles. These examples show that recombination in itselfdoes not cause homogenization of introns among all alleles at a locus. Rather,over evolutionary time, genetic drift will lead to homogenization of introns,given that recombination occurs.

Ann

u. R

ev. G

enet

. 199

8.32

:415

-435

. Dow

nloa

ded

from

arj

ourn

als.

annu

alre

view

s.or

gby

b-o

n: U

nive

rsid

ade

do P

orto

(U

P) o

n 02

/10/

09. F

or p

erso

nal u

se o

nly.



432 HUGHES & YEAGER

Bergstrom and colleagues (4) recently sequenced portions of the second in-tron from the human class IIDRB1locus. They found that the intron sequenceswere much more similar to each other than were sequences of exon 2 (whichincludes the PBR); indeed, intron sequences were even more similar than weresynonymous sites in exons. These results are very similar to the class I results,which suggests that introns are younger than exons because introns have beenhomogenized relative to exons by recombination and subsequent genetic drift.However, Bergstrom et al (4) favor a different interpretation of their data. Theyargue that in fact the introns reveal the true age of the MHC alleles. Thus,according to these authors, allelic lineages are not really as old as predictedunder the hypothesis of trans-species polymorphism. Rather, they argue,DRB1alleles are really very recent.

Bergstrom et al (4) do not address the fact that their conclusions contradictthose of previous studies that suggested that MHC alleles are very ancient.Moreover, if they are right, the class II exons have experienced very rapid re-cent evolutionat synonymous sites. Positive selection will increase the rate ofnucleotide substitution at nonsynonymous sites, but there is no known mecha-nism that will increase the rate of substitution at synonymous sites. Rather, weexpect the rate of substitution at synonymous sites to be very similar to that atsites in introns; and this prediction is supported by comparison of mammalianintrons and exons in the case of non-MHC genes (28). In spite of their inter-pretation, the Bergstrom et al (4) data clearly suggest thatDRB1exon 2 se-quences are much older thanDRB1intron 2 sequences—as seen in class I andas predicted by population genetics theory (10).

It is important to distinguish the effects of hitch-hiking of one locus withanother, linked locus in two different circumstances: (a) when the linked locusis under directional selection (which leads to fixation of a selectively favoredallele); and (b) when the linked locus is under balancing selection. In the formercase, both the locus under selection and a closely linked locus will show re-duced polymorphism compared to neutral loci because of the recent fixation ofthe favorable allele. This phenomenon is often referred to as a selective sweep,because polymorphism linked to the favored mutation is swept out of the popu-lation as the new mutant goes to fixation. By contrast, a locus closely linked to alocus under balancing selection will show higher polymorphism than a neutrallocus (43, 54). The polymorphism seen at such a linked locus will be a functionof how tightly it is linked to the selected locus. In the case ofDRB1, the selectionis acting on nonsynonymous sites in the PBR codons in exon 2. Because sitesin intron 2 ofDRB1are less closely linked to PBR nonsynonymous sites thanare synonymous sites in exon 2, the latter are expected to show a higher levelof polymorphism because of their hitch-hiking with the PBR nonsynonymoussites, exactly as theDRB1data show (4).

Ann

u. R

ev. G

enet

. 199

8.32

:415

-435

. Dow

nloa

ded

from

arj

ourn

als.

annu

alre

view

s.or

gby

b-o

n: U

nive

rsid

ade

do P

orto

(U

P) o

n 02

/10/

09. F

or p

erso

nal u

se o

nly.




ACKNOWLEDGMENTS

This research was supported by grants R01-GM34940 and K04-GM00614 fromthe National Institutes of Health.

Visit the Annual Reviews home pageathttp://www.AnnualReviews.org

Literature Cited

1. Baily DW, Kohn HI. 1965. Inherited histo-compatibility changes in progeny of irra-diated and unirradiated mice.Genet. Res.6:330–40

2. Belich MP, Glynne RY, Senger G, SheerD, Trowsdale J. 1994. Proteasome com-ponents with reciprocal expression to thatof the MHC-encoded LMP proteins.Curr.Biol. 4:769–76

3. Bender BS, Crogham T, Zhang L, SmallPA Jr. 1992. Transgenic mice lackingclass I major histocompatibility complexrestricted T cells have delayed viral clear-ance and increased mortality after in-fluenza virus challenge.J. Exp. Med.175:1143–45

4. Bergstrom TF, Josefsson A, Erlich HA,Gyllensten UB. 1997. Analysis of intronsequences at the class II HLA-DRB1 locus:implications for the age of allelic diversity.Hereditas127:1–5

5. Berke G. 1994. The binding and lysis of tar-get cells by cytotoxic lymphocytes: molec-ular and cellular aspects.Annu. Rev. Im-munol.12:735–73

6. Bjorkman PJ, Saper MA, Samraoui B, Ben-net WAS, Strominger JL, Wiley DC. 1987.Structure of the human class I histocompat-ibility antigen, HLA-A2.Nature329:506–12

7. Bjorkman PJ, Saper MA, Samraoui B, Ben-net WAS, Strominger JL, Wiley DC. 1987.The foreign antigen binding site and T cellrecognition regions of class I histocompat-ibility antigens.Nature329:512–18

8. Brown JH, Jardetzky T, Saper MA, Sam-raoui B, Bjorkman PJ, Wiley DC. 1988. Ahypothetical model of the foreign antigenbinding site of class II histocompatibilitymolecules.Nature332:845–50

9. Brown JH, Jardetzky T, Saper MA, Sam-raoui B, Bjorkman PJ, Wiley DC. 1993.Three-dimensional structure of the humanclass II histo-compatibility antigen HLA-DR1.Nature364:33–39

10. Cereb N, Hughes AL, Yang SY. 1997.Locus-specific conservation of theHLA

class I introns by intra-locus homogeniza-tion. Immunogenetics47:30–36

11. Clarke AG. 1971. The effects of mater-nal pre-immunization on pregnancy in themouse.J. Reprod. Fertil.24:369–75

12. Clarke B, Kirby DR. 1966. Maintenanceof histocompatibility complex polymor-phisms.Nature211:999–1000

13. Doherty PC, Zinkernagel RM. 1975. En-hanced immunologic surveillance in miceheterozygous at theH-2gene complex.Na-ture256:50–52

14. Driscoll J, Finley D. 1992. A controlledbreakdown: antigen processing and theturnover of viral proteins.Cell 68:823–25

15. Felsenstein J. 1985. Confidence limits onphylogenies: an approach using the boot-strap.Evolution39:783–91

16. Flynn JL, Goldstein MM, Triebold KJ,Koller B, Bloom BR. 1992. Major his-tocompatibility complex class I-restrictedT cells are required for resistance toMy-cobacterium tuberculosisinfection. Proc.Natl. Acad. Sci. USA89:12013–17

17. Germain RN. 1994. MHC-dependent anti-gen processing and peptide presentation:providing ligands for T-lymphocyte acti-vation.Cell 76:287–99

18. Guo H-C, Jardetzky TS, Garrett TPJ, LaneWS, Strominger JL, Wiley DC. 1992. Dif-ferent length peptides bind to HLA-Aw68similarly at their ends but bulge out in themiddle.Nature360:364–66

19. Gyllensten UB, Erlich HA. 1989. Ancientroots for polymorphism at the HLA-DQAlocus in primates.Proc. Natl. Acad. Sci.USA86:9986–90

20. Hedrick PW, Black FL. 1997. HLA andmate selection: no evidence in SouthAmerindians.Am. J. Hum. Genet.61:494–96

21. Hedrick PW, Thomson G. 1983. Evidencefor balancing selection at HLA.Genetics104:449–56

22. Hughes AL. 1997. Evolution of the pro-teasome components.Immunogenetics46:82–92

Ann

u. R

ev. G

enet

. 199

8.32

:415

-435

. Dow

nloa

ded

from

arj

ourn

als.

annu

alre

view

s.or

gby

b-o

n: U

nive

rsid

ade

do P

orto

(U

P) o

n 02

/10/

09. F

or p

erso

nal u

se o

nly.



434 HUGHES & YEAGER

23. Hughes AL, Hughes MK. 1995. Natu-ral selection on the peptide-binding re-gions of major histocompatibility complexmolecules.Immunogenetics42:233–43

24. Hughes AL, Nei M. 1988. Pattern of nu-cleotide substitution at major histocompat-ibility complex class I loci reveals over-dominant selection.Nature335:167–70

25. Hughes AL, Nei M. 1989. Nucleotide sub-stitution at major histocompatibility com-plex class II loci: evidence for overdomi-nant selection.Proc. Natl. Acad. Sci. USA86:958–62

26. Hughes AL, Nei M. 1989. Evolution of themajor histocompatibility complex: inde-pendent origin of nonclassical class I genesin different groups of mammals.Mol. Biol.Evol.6:559–79

27. Hughes AL, Nei M. 1990. Evolutionary re-lationships of class II MHC genes in mam-mals.Mol. Biol. Evol.7:491–514

28. Hughes AL, Yeager M. 1997 Comparativeevolutionary rates of introns and exons inmurine rodents.J. Mol. Evol.45:125–30

29. Hughes AL, Hughes MK, Howell CY, NeiM. 1994. Natural selection at the class IImajor histocompatibility complex loci ofmammals.Phil. Trans. R. Soc. London Ser.B 345:359–67

30. James DA. 1965. Effects of antigenic dis-similarity between mother and foetus onplacental size in mice.Nature 205:613–14

31. James DA. 1967 Some effects of immuno-logical factors on gestation in mice.J. Re-prod. Fertil.14:265–75

32. Klein J. 1986.Natural History of the MajorHistocompatibility Complex. New York:Wiley

33. Kreitman M, Hudson RR. 1991. Inferringthe evolutionary histories of theAdh andAdh-duploci in Drosophila melanogasterfrom patterns of polymorphism and diver-gence.Genetics127:565–82

34. Kropshoffer H, Hammerling GJ, Vogt AB.1997. How HLA-DM edits the MHCclass II peptide repertoire: survival of thefittest?Immunol. Today18:77–82

35. Lawlor DA, Ward FF, Ennis PD, JacksonAP, Parham P. 1988. HLA-A,-B polymor-phisms predate the divergence of humansand chimpanzees.Nature335:268–71

36. Lopez de Castro JA, Strominger JL, StrongDM, Orr HT. 1982. Structure of crossre-active human histocompatibility antigensHLA-A28 and HLA-A2: possible impli-cations for the generation of HLA poly-morphism. Proc. Natl. Acad. Sci. USA79:3813–17

37. Maruyama T, Nei M. 1981. Genetic vari-ability maintained by mutation and over-

dominant selection in finite populations.Genetics98:441–59

38. Mayer WE, Jonker D, Klein D, Ivanyi P,van Seventer G, Klein J. 1988. Nucleotidesequence of chimpanzee MHC class I al-leles: evidence for trans-species mode ofevolution.EMBO J.7:2765–74

39. Monaco JJ. 1992. A molecular model ofMHC class-I-restricted antigen processing.Immunol. Today13:173–78

40. Nathenson SG, Geliebter J, PfaffenbachGM, Zaff RA. 1986. Murine major his-tocompatibility complex class-I mutants:molecular analysis and structure-functionimplications.Annu. Rev. Immunol.4:471–502

41. Nei M, Gojobori T. 1986. Simple methodsfor estimating the number of synonymousand nonsynonymous nucleotide substitu-tions.Mol. Biol. Evol.3:418–26

42. Nei M, Jin L. 1989. Variances of the av-erage numbers of nucleotide substitutionswithin and between populations.Mol. Biol.Evol.6:290–300

43. Nei M, Li W-H. 1980. Non-random associ-ation between electromorphs and inversionchromosomes in finite populations.Genet.Res.35:65–83

44. Ober C, Weitkamp LR, Cox N, Dytch H,Kostyu D, Elias S. 1997. HLA and matechoice in humans.Am. J. Hum. Genet.61:497–504

45. Ohta T. 1982. Allelic and nonallelic ho-mology of a supergene familyProc. Natl.Acad. Sci. USA79:3251–54

46. Orlowski M. 1990. The multicatalytic pro-teinase complex: a major extralysoso-mal proteolytic system.Biochemistry29:10289–97

47. Peters PJ, Neefjes JJ, OOrschot V, PloeghHL, Geuze HJ. 1991. Segregation of MHCclass II molecules from MHC class Imolecules in the Golgi complex for trans-port to lysosomal compartments.Nature349:669–76

48. Potts WK, Manning CJ, Wakeland EK.1991. Mating patterns in seminatural pop-ulations of mice influenced by MHC geno-type.Nature352:619–21

49. Rammensee H-G, Friede T, StevanovicS. 1995. MHC ligands and peptide mo-tifs: first listing. Immunogenetics41:178–228

50. Rivett AJ. 1993. Proteasomes: multicat-alytic proteinase complexes.Biochem. J.291:1–10

51. Saitou N, Nei M. 1987 The neighbor-joining method: a new method for re-constructing phylogenetic trees.Mol. Biol.Evol.4:406–25

52. Saper MA, Bjorkman PJ, Wiley DC. 1991.

Ann

u. R

ev. G

enet

. 199

8.32

:415

-435

. Dow

nloa

ded

from

arj

ourn

als.

annu

alre

view

s.or

gby

b-o

n: U

nive

rsid

ade

do P

orto

(U

P) o

n 02

/10/

09. F

or p

erso

nal u

se o

nly.




Refined structure of the human histocom-patibility antigen HLA-A2 at 2.6A resolu-tion. J. Mol. Biol.219:277–319

53. Silver ML, Guo H-C, Strominger JL, Wi-ley DC. 1992. Atomic structure of a humanMHC molecule presenting an influenzavirus peptide.Nature360:367–69

54. Strobeck C. 1983. Expected linkage dise-quilibrium for a neutral locus linked to achromosomal arrangement.Genetics103:545–55

55. Takahata N, Nei M. 1990. Allelic geneal-ogy under overdominant and frequency-dependent selection and polymorphismof major histocompatibility complex loci.Genetics124:967–78

56. Thomas L. 1974. Biological signals for

self-identification. InProgress in Immunol-ogy II, ed. L Brent, J Holborrow, pp. 239–47. Amsterdam: North-Holland

57. Wegmann TG. 1984. Foetal protectionagainst abortion: is it immuno-suppressionor immunostimulation? Ann. Immunol.135B:307–12

58. Yamazaki K, Boyse EA, Mike V, ThalerHT, Mathieson BJ, et al. 1976. Control ofmating preferences in mice by genes in themajor histocompatibility complex.J. Exp.Med.114:1324–35

59. Zinkernagel RM, Doherty PC. 1974. Im-munological surveillance against alteredself components by sensitized T lympho-cytes in lymphocytic choriomeningitis.Na-ture251:547–48

Ann

u. R

ev. G

enet

. 199

8.32

:415

-435

. Dow

nloa

ded

from

arj

ourn

als.

annu

alre

view

s.or

gby

b-o

n: U

nive

rsid

ade

do P

orto

(U

P) o

n 02

/10/

09. F

or p

erso

nal u

se o

nly.

Annual Review of Genetics Volume 32, 1998

CONTENTSAlfred D. Hershey, Allan Campbell, Franklin W. Stahl 1

The Role of the FHIT/FRA3B Locus in Cancer , Kay Huebner, Preston N. Garrison, Larry D. Barnes, Carlo M. Croce 7

Regulation of Symbiotic Root Nodule Development, M. Schultze, A. Kondorosi 33

Targeting and Assembly of Periplasmic and Outer-Membrane Proteins in Escherichia coli , Paul N. Danese, Thomas J. Silhavy 59

The Genetics of Breast Cancer Susceptibility, Nazneen Rahman, Michael R. Stratton 95

Nonsegmented Negative Strand RNA Viruses: Genetics and Manipulation of Viral Genomes, Karl-Klaus Conzelmann 123

The Genetics of Disulfide Bond Metabolism, Arne Rietsch, Jonathan Beckwith 163

Comparative DNA Analysis Across Diverse Genomes, Samuel Karlin, Allan M. Campbell, Jan Mrázek 185

The Ethylene Gas Signal Transduction Pathway: A Molecular Perspective, Phoebe R. Johnson, Joseph R. Ecker 227

Molecular Mechanisms of Bacteriocin Evolution, Margaret A. Riley 255

Alternative Splicing of Pre-mRNA: Developmental Consequences and Mechanisms of Regulation, A. Javier Lopez 279

Kinetochores and the Checkpoint Mechanism that Monitors for Defects in the Chromosome Segregation Machinery, Robert V. Skibbens, Philip Hieter

307

The Diverse and Dynamic Structure of Bacterial Genomes, Sherwood Casjens 339

Recombination and Recombination-Dependent DNA Replication in Bacteriophage T4, Gisela Mosig 379

Natural Selection at Major Histocompatibility Complex Loci of Vertebrates, Austin L. Hughes, Meredith Yeager 415

Evolution and Mechanism of Translation in Chloroplasts, Masahiro Sugiura, Tetsuro Hirose, Mamoru Sugita 437

Genetics of Alzheimer's Disease, Donald L. Price, Rudolph E. Tanzi, David R. Borchelt, Sangram S. Sisodia 461

THE CRITICAL ROLE OF CHROMOSOME TRANSLOCATIONS IN HUMAN LEUKEMIAS, Janet D. Rowley 495

Early Patterning of the C. elegans Embryo, Lesilee S. Rose, Kenneth J. Kemphues 521

Ann

u. R

ev. G

enet

. 199

8.32

:415

-435

. Dow

nloa

ded

from

arj

ourn

als.

annu

alre

view

s.or

gby

b-o

n: U

nive

rsid

ade

do P

orto

(U

P) o

n 02

/10/

09. F

or p

erso

nal u

se o

nly.

Genetic Counseling: Clinical and Ethical Challenges, M. B. Mahowald, M. S. Verp, R. R. Anderson 547

Mating-Type Gene Switchingin Saccharomyces cerevisiae , James E. Haber 561

Epitope Tagging, Jonathan W. Jarvik, Cheryl A. Telmer 601

The Leptotene-Zygotene Transition of Meiosis, D. Zickler, N. Kleckner 619

Ann

u. R

ev. G

enet

. 199

8.32

:415

-435

. Dow

nloa

ded

from

arj

ourn

als.

annu

alre

view

s.or

gby

b-o

n: U

nive

rsid

ade

do P

orto

(U

P) o

n 02

/10/

09. F

or p

erso

nal u

se o

nly.

NATURAL SELECTION AT MAJOR HISTOCOMPATIBILITY...

Documents

Transcript of NATURAL SELECTION AT MAJOR HISTOCOMPATIBILITY...